Should you tell clients you use AI to build their website?
Should a one-person AI agency tell clients it uses AI to build their website? Yes, plainly and early. What clients worry about, what to say, how to price it.
A web designer who runs a one-person AI agency is on a call with a café owner who wants a new website before the summer menu launches. It's going well. Then the owner says, half joking, "You're not going to get a robot to write it, are you?"
The designer does use an AI agent. It drafts pages, writes code and suggests wording. She decides what the site needs, reads every line, and fixes what's wrong. The question is how to say that without the call turning awkward.
Tell them. Say it early, in plain words, and put the weight on what you do: you decide, you check, you're responsible. The worry behind the joke is rarely the tool itself. It's being surprised later, or paying for something nobody looked at.
Should I tell clients I use AI to build their site?
Yes, for three practical reasons.
They'll find out anyway. A client who hears it from you on the first call hears a method. A client who spots it later hears a secret.
It sets up the right conversation. Once it's said, you can talk about the part that matters to them: how you check the work and what happens if something's wrong.
It protects you. If a client later feels misled, the site's quality stops mattering. A sentence in your proposal about how you work costs nothing and avoids that argument.
The way you say it matters more than whether you say it. "I use AI" on its own invites worry. "I build with an AI agent, and I check every page myself before you see it" answers the worry in the same breath.
What do clients worry about?
Listen for the worry under the question. It's usually one of these:
| What they ask | What they're worried about | What helps |
|---|---|---|
| "Will it look like everyone else's?" | A generic site that doesn't feel like their business | Show how you use their photos, their words and their customers' questions |
| "Will Google mark it down?" | Losing customers from search | Explain Google's published position (below) and how you check facts |
| "Is it just made up?" | Wrong prices, wrong hours, invented claims | Walk through how you check every fact against what they gave you |
| "Why am I paying you, then?" | Being overcharged for a machine's work | Talk about the result, the checking and who fixes it when it breaks |
| "What happens to my information?" | Their details, or their customers', ending up somewhere public | Say what you share with the tool and what you never share |
Each of these has a straight answer. None of them needs you to hide how you work.
What does Google say about AI-built websites?
Google published its position in February 2023, and it's plain enough to quote to a client. Its ranking systems aim to reward original, helpful, high-quality content, however it's produced. Using automation, including AI, mainly to manipulate search rankings is against its spam policies. Google's own FAQ on that page says appropriate use of AI isn't against its guidelines.
So the honest answer to "Will Google mark it down?" is: Google looks at whether the site is helpful and accurate, not at which tool typed it. A site full of the café's real menu, real photos and real opening hours is exactly what that guidance describes. A site stuffed with pages written only to catch searches is not, whoever wrote it.
Don't turn this into a promise about rankings. Nobody can promise where a site will appear. You can promise that it will be accurate, useful and checked.
How do I explain the way I work?
Describe the steps, and make it clear where you stand in each one. A short version for a proposal or a first call:
- I learn the business. I ask about your customers, your services and what you want the site to do.
- I build with an AI agent. It drafts pages and writes the code, working from the notes and facts you've given me.
- I check everything. Every page, every fact, every link and form, on a phone and a computer.
- You approve it. You see the exact version before it goes live.
- I stay responsible. If something's wrong, I fix it.
Step three is where your value is, so be specific about it. A client who hears that you test every form and check every price on a real phone understands what they're paying for. The checklist in how to check a website before handover is a good thing to describe, or even share.
If you'd like to see the build side in more detail, the walkthrough on using Claude Code to build a website for a client covers how the agent fits into a real project.
What does that sound like on a call?
Say it before they ask, and keep it short. Here is how the café call could go:
"I build with an AI agent. It does the drafting and the code, which means more of your budget goes on getting the details right. I read and check every page myself, and you'll approve the final version before it goes live. If anything's wrong after launch, I fix it."
That's four sentences. It names the tool, says what it's for, says what you do, and says who's responsible. That is often enough to move the call back to the menu.
If they don't, ask what worries them. You'll usually find one of the concerns in the table above, and each has a straight answer.
If AI makes it faster, should I charge less?
That's your decision, but don't let the tool decide it for you. The client isn't buying your hours. They're buying a site that works, says the right things, and is looked after when something breaks.
Faster building gives you options. You can spend the saved time on the parts clients notice (better photos, clearer service pages, a proper test on every phone) and keep the price. You can lower the price and take on more projects. Or you can offer a smaller starting package for businesses with a tight budget. Any of these is honest, as long as the quote describes what they get.
What isn't honest is charging for time you didn't spend. If you price by the hour, count real hours.
What should I say about their information?
Be specific, because vague reassurance sounds like hiding something. Tell the client:
- what you give the AI tool: the brief, the business facts, the copy you've agreed;
- what you never give it: passwords, and their customers' private details;
- that you check what it writes before it goes anywhere near their site.
Then do exactly that. Keep logins in a password manager, not in notes the agent reads.
Who approves what the AI writes?
You do, and then the client does. Nothing the agent drafts should reach the client, their customers or the live site until you've read it. That habit is what makes the rest of this conversation true.
It helps to have the step built into your tools, so it doesn't depend on remembering. Approving what your agent sends shows one way to keep it: you see the exact page or message, and it goes out only when you say yes. For the full process from brief to launch, building, reviewing and delivering client websites covers the stages in order.
If you're a developer building client sites this way, the page for developers building client sites with an AI agent shows how the rest of the working week fits around it.
The short version for the café owner: yes, I use AI; no, it doesn't decide anything; and I'm the one who answers for the site.